Assessing the Effect of NAFTA's Rules of origin
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Assessing the Effect of NAFTA’s Rules of origin*
Olivier Cadot (Université de Lausanne, CERDI and CEPR)
Jaime de Melo (Université de Genève, CERDI and CEPR)
Antoni Estevadeordal (IADB)
Akiko Suwa-Eisenmann (INRA and DELTA)
Bolormaa Tumurchudur (Université de Lausanne)
Revised version,
June 2002
1. Introduction
In the last decade or so, a voluminous literature (see e.g. de Melo and Panagarya, 1992;
Bhagwati and Panagaryia, 1996; Frankel, 1997; or Bhagwati, Krishna and Panagariya, 1999,
and references therein) has attempted to assess the implications of Preferential Trading
Arrangements (PTA) for trade patterns, global welfare, and the multilateral trading system. If
this literature has yet to converge to an unambiguous answer as to whether PTAs are, in a
second-best world, good or bad for world welfare, the theoretical arguments are by now fairly
clear (see e.g. Bhagwati and Panagariya, 1999). By contrast, the empirical evidence is still
scant. In particular, one basic question remains largely unanswered: to what extent do PTAs
really improve market access for member countries, in particular for smaller ones in
asymmetric blocs? The answer is less obvious than it looks. Notwithstanding the fact that
GATT Article XXIV, para. 8 (b) requires the removal of tariff barriers on “substantially all
trade” inside Free-Trade Areas, non-tariff measures can create significant barriers to intra-
bloc trade.
*
Cadot, de Melo and Suwa gratefully acknowledge financial support from the IBRD. Special thanks to Jose
Anson for superb research assistance, and to Alessandro Nicita and Marcelo Olarreaga for providing us with
data. We are also grateful to Daniel Lederman, Marcelo Olarreaga, three anonymous referees and participants at
the NAFTA Workshop held in Washington, DC in May 2002 for useful comments. The positions expressed in
this paper do not reflect the views of the institutions to which the authors are affiliated, and any remaining errors
remain our entire responsibility..
1One of the non-tariff measures most likely to act as a barrier to internal trade is the array of
rules of origin (ROO) that accompany virtually all PTAs. The general idea of ROOs is (i) to
prevent the trans-shipment of goods imported into the area via member states with low
external tariffs into member states with higher ones; (ii) to prevent the superficial
transformation of non-originating goods in one member state for re-export into others. Such
rules (sometimes reflecting capture by high-cost producers in member states with strong
bargaining positions) can entail large compliance costs for intra-bloc exporters forced to
source intermediate inputs in the area when outside suppliers would be cheaper.
The direct cost of complying with ROOs (essentially higher input prices) is often
compounded by administrative or bookkeeping costs. For instance, using firm-level data,
Koskinen (1983) estimated administrative compliance costs under the EFTA-EC FTA at
between 1.4% and 5.7% of the value of export transactions, while according to Holmes and
Shephard (1983) the average export transaction from EFTA to the EC required 35 documents
and 360 copies.1 The combination of bookkeeping costs with ROO-induced constraints on
international sourcing can be particularly penalizing for companies operating globally
integrated supply chains, which is now the norm in many manufacturing sectors.
Because of such compliance costs, as noted by a recent UNCTAD document, “[t]he mere
granting of tariff preferences or duty-free market access to exports originating in LDCs does
not automatically ensure that the trade preferences are effectively utilized by beneficiary
countries” (UNCTAD 2001, p. 8). This syndrome, referred to in the same document as
“missed preferences” --preferential treatment that exporters choose not to use-- affects most
preferential trading arrangements.2 Of course, a PTA offering no market-access gains at all
would hardly attract any members, but in the case of asymmetric PTAs like NAFTA or the
EU’s Europe Agreements with the CEECs, political benefits for smaller member states
(credibility, anchoring of reforms, and so on) can be large enough to make joining attractive
even when direct trade benefits are small. Thus, a full analysis of the market-access
implications of PTAs for member states should look at both tariff preferences and compliance
costs. This is what the present paper sets out to do for NAFTA.
1
Quoted in Herin (1986)
2
In 1999, for instance, whereas the EU's GSP theoretically covered 99% of EU imports from eligible countries,
only 31% was utilized by exporters in those countries (Brenton and Manchin 2002).
2Whereas a few pioneering studies (see e.g. Herin 1986, Estevadeordal 1999 or Estevadeordal
and Miller 2002) have attempted to assess empirically the effect of ROOs, the evidence is still
hard to come by. One basic reason has to do with measurement: The term “rules of origin”
covers a wide range of technical measures codified in voluminous legal documents like
NAFTA’s Annex 401. Translating legally-expressed technical measures into a variable or
vector of variables that can be mapped into dollar costs via a well-defined mathematical
function is a daunting task.
In this paper, we approach the problem of ROO measurement in two distinct ways. First,
following Herin (1986), we estimate indirectly the cost of complying with ROOs via a
revealed-preference mechanism, using data on NAFTA utilization rates provided by the ITC.
The choice facing Mexican exporters is to send goods under NAFTA, which entitles them to
tariff preferences but requires compliance with NAFTA’s rules, in particular ROOs, or to
send them under MFN, which forces them to pay the US’s MFN tariff but does not require
compliance with NAFTA rules. In sectors where NAFTA’s utilization rate is 100%, the
benefit of tariff preference is revealed larger than compliance costs. Thus, in those sectors, the
rate of tariff preference provides an upper bound on the ad-valorem equivalent of compliance
costs. In sectors where NAFTA’s utilization rate is zero, by the same reasoning, tariff
preference provides a lower bound on compliance costs. Finally, in all sectors where
utilization rates are strictly between zero and 100%, if Mexican exporters had symmetric
compliance costs they would be revealed indifferent between shipping under NAFTA or
MFN, so the rate of tariff preference would be revealed equal to the compliance costs. With
heterogenous compliance costs, all that can be said is that the tariff preference gives a rough
estimate of compliance costs in the sense that at least some firms have higher compliance
costs whereas some have lower ones.
Second, we use an index of ROO strictness constructed by Estevadeordal (2000) using direct
information from NAFTA's Annex 401. We use this index to estimate the combined effects of
NAFTA’s tariff preference and ROOs on the direction of Mexican exports. If the tariff
preference substantially improved Mexican access to the US market, one would expect this to
affect positively the latter’s share in Mexican exports. By the same token, one would expect
ROOs to have the opposite effect. Estimating these two effects in one equation gives a rough
estimate of the magnitudes of the two conflicting effects. Indeed, using trade data at the 6-
3digit level we show that tariff preference has a positive and highly significant effect on the
direction of Mexican exports whereas ROOs have a negative and similarly significant effect.
Moreover, simple simulations suggest that these effects roughly balance one another, so that
the net effect of NAFTA treatment on Mexican export direction was of small magnitude. This
result is consistent with the fact that a substantial chunk of Mexico’s exports to the US enter
with NAFTA utilization rates below 100%, and also with evidence reported by Krueger
(1999) suggesting that NAFTA had only modest effects on Mexican trade flows.3 In the same
vein, using a gravity equation with dummy variables for bloc membership Soloaga and
Winters (1999) found that “[…] it seems that the key developments in NAFTA members’
trade policies (Mexico’s unilateral liberalization in mid-80’s, CUSFTA in 1988 and NAFTA
itself signed by the end of 1992) were not associated with appreciable changes in intra or
extra-bloc trade, once we take into account the ‘normal’ variation in trade levels that follows
changes in the gravity variables.” (p. 12) Not only are our results consistent with those of
Krueger and Soloaga and Winters; they also point to rules of origin as the prime culprit,
confirming widely held suspicions and anecdotal evidence.
This has potentially important implications for the FTAA currently on the agenda of the US
administration. Extending NAFTA Southward would in principle carry several advantages.
First, the literature on PTAs has stressed the benefits of “open regionalism”, and Chile’s long-
standing efforts to sign trade agreements with NAFTA’s members show that trade integration
across the continent has perceived benefits even for geographically remote countries. Second,
bilateral agreements such as Chile’s FTA with Canada create a hub-and-spoke system with
only limited potential for market-access gains, especially in the presence of rules of origin.
Extending NAFTA to the South, as Wonnacott (1996) pointed out, has stronger potential to
generate market-access gains for new members. Finally, unlike Mercosur or the European
Community, NAFTA (being an FTA) does not force new members to align their levels of
external protection, which is an advantage for countries like Chile that have already reduced
their trade barriers unilaterally.
However, in what can undoubtedly become what Bhagwati and Panagariya (1996) called a
“spaghetti bowl” of criss-crossing PTAs, ROOs may limit considerably the potential for
3
However Krueger (1999, Table 3) did find action in capital-good industries (SITC 7, machinery and transport
equipment) where Mexican exports to the US increased two-fold between 1990 and 1998 through a substantial
increase in Mexican market share.
4market-access improvement. In a CU, by contrast, the CET eliminates the risk of trade
deflection, reducing the need for ROOs whose sole purpose is then the protection of domestic
content (which can nevertheless be an important motive by itself). This prompted Wonnacott
(1996) to suggest that a CU or some sort of hybrid arrangement could be better than a web of
overlapping FTAs with complex ROOs. Thus, whether or not ROOs do constitute a real
barrier to market access is a key question for the design of a pan-American trade agreement,
and indeed we show that, given the pattern of NAFTA’s ROOs and current Latin American
trade flows, prospects for spectacular market-access improvements through NAFTA-like
treatment appear limited.
The paper is organized as follows. In section 2, we review background issues. We first derive
basic formulas for ROOs highlighting their relationship with effective-protection formulas,
then give a brief description of NAFTA’s ROO regimes and describe the empirical
relationship between ROOs and tariff preference. In section 3, we provide revealed-preference
estimates of the cost of ROOs and analyse simultaneously the effects of tariff preferences and
ROOs on the direction of Mexican exports. In section 4, we draw some tentative lessons for
the likely effects of extending NAFTA-like treatment to other Latin American countries under
the Free Trade Agreement of the Americas (FTAA). Finally, section 5 concludes.
2. Basic issues
2.1 ROOs and effective protection
Suppose that a final good, denoted by the subscript F, is produced with labor (l) and an
intermediate good, denoted by the subscript I, according to the following technology:
xF = min { f(lF); (xI+xI*)/a} (1)
where a is an input-output coefficient, xI represents the quantity of intermediate originating
from the free-trade area (henceforth called “home-made”) and xI* the quantity imported from
the rest of the world. The formulation in (1) presupposes that the home-made and imported
varieties of the intermediate good are perfect substitutes in production, an assumption that we
will maintain throughout. Technology f produces value added with labor under decreasing
5returns to scale, due to the presence of a fixed factor omitted from the formulation. Two
countries, North (N) and South (S) form a Free-Trade Agreement (FTA) whereby goods
produced in one of the two countries can be exported to the other at preferential (reduced)
tariff rates provided that they satisfy ROO. In order to keep things simple, we will suppose
that the South does not produce the intermediate good and hence does not protect it. Let pF*
and tF be respectively the final good's world price and ad-valorem external tariff in the North.
Let also pI* and pI be the intermediate good's world price and domestic price in the free-trade
area. The latter is determined endogenously by the ROO.
A regional value content (RVC) expressed by weight can be modelled as a constraint on the
quantity (volume) of the imported variety that can be used by the final-good producer in order
to qualify for preferential treatment. Suppose that the ROO specifies that a proportion α of
intermediate consumption must originate from the area. Formally, the rule takes the form
xI ≥ α(xI + xI*). (2)
If the Northern intermediate-good industry is inefficient at world prices, the ROO is binding,
in which case (2) can be written as
α *
xI = xI . (3)
1−α
Using this, the total value of intermediate consumption at domestic prices is
α p
ϕ (α ) = p I x I + p I x I* = p I + p I* x I* = I x I* (4)
1 − α 1−α
where p I = αp I + (1 − α ) p I * . The volume of intermediate consumption is
α * 1 *
x I + x I* = 1 + xI = xI . (5)
1−α 1−α
Note that pI is the price index of the composite intermediate good given the ROO: This can
be seen by dividing (4) by (5). Because the ROO segments the intermediate-good market in
the free-trade area, the price of the home-made one is determined endogenously by supply
and demand. Letting xIN(pI) be its supply (in the North, since the South does not produce it)
and xI (pI, pF* (1+tF), α) its demand. Demand for the home-made intermediate comes only
6from the South, since Northern producers won't bother to use it when its price is higher than
pI*(1+tI). That price is determined by4
xI (,pI, pF*(1+tF), α) = xIN (pI) (6)
Suppose first that Southern producers of the final good enjoy tariff-free access to the Northern
market if they satisfy the ROO. Unit value added is then
v(tI,α) = pF*(1+tF) – a pI
whereas, under MFN regime, v*=pF* - apI*. The benefit from using the preferential regime
is thus
b = v(tI,α) - v* = pF*tF – a ( pI - pI*) = pF*tF - α a (pI - pI*). (7)
Let ρI = (pI - pI*)/ pI* be the ad-valorem equivalent (AVE) of the premium on the home-made
intermediate good generated by the ROO. Then (7) can be rewritten as
b = pF*tF – α apI* ρI (8)
When α = 1, (8) boils down to an effective-protection expression and b is positive whenever
the effective rate of protection on final-good production, given by a tariff tF on the final good
and a tariff equivalent ρI on the intermediate good, is positive. When α < 1, the preferential
regime can be profitable (b > 0) even when the rate of effective protection given tariff
equivalents tF and ρI is negative. This is because α t −τ F
b = p*F F − aα ( p I − p*I ) (9)
1+τ F
It follows directly from (9) that b is decreasing in τF, (hence increasing in the rate of
preference) and decreasing in α, the strictness of the ROO.
Suppose now that the terms of PTAs (tariff preferences and rules of origin) are set so as to
leave partners with weak bargaining power close to or on their “participation constraint” (a
term borrowed from contract theory meaning “just indifferent between signing and not
signing”). In NAFTA’s case, unlike in Lome-type arrangements, the Southern partner gave
substantial tariff concessions, so that its participation constraint must imply at least some
market-access gains in the North. However, NAFTA is sufficiently asymmetric to admit the
idea that Mexico's participation constraint implied a relatively low level of market-access
concessions from the US. Moreover, as argued above, non-trade gains may have been
substantial for the Mexican government which was, at the time of NAFTA’s formation,
engaged in substantial reforms in need of political anchoring. Thus, the idea of Mexico being
by and large on its participation constraint when NAFTA was formed is a plausible one.
Returning to the algebra, this means that the value of b can be set identically equal to zero in
(9). The right-hand side of the equation can then been differentiated totally with respect to in
τF and α, giving dα/dτF < 0: deeper preferences (a lower value of τF) are associated with
stricter ROOs (a higher value of α).
Thus, the participation-constraint approach suggests substitutability between tariffs and ROOs
as instruments of intra-bloc protection. What, then determines the use of either one? A full
answer would require that we set up a political-economy model, which is beyond the scope of
this paper. But note that, in the stripped-down framework above, if tariffs and ROOs are
perfect substitutes in the Northern member’s objective function, even in the presence of a
participation constraint the optimal instrument mix is indeterminate. Thus, the answer must
have to do with tariffs and ROOs having distinct distributional implications. Suppose that
Southern final-good producers used to procure their intermediate inputs outside of the bloc
before the preferential agreement. As ROOs are substituted for tariff protection, they start
the price satisfying the market-equilibrium condition is above pI*(1+tI).
8procuring in the bloc, say in the North if the South does not produce the intermediate inputs.
As a result, their costs go up and they enjoy only marginally improved market access. But
Northern producers of the intermediate good now enjoy a captive market and emerge as the
winners. Thus, Northern (US) intermediate-good producers can be expected to lobby in
favour of ROOs. Taking the argument one step further, as stiffer ROOs require deeper tariff
preferences along the Southern member’s participation constraint (because dα/dτF < 0 along b
= 0, see supra), Northern intermediate-good producers can be expected to lobby in favour of
deeper tariff preferences in their downstream sectors. When they succeed, tariff revenue for
the Northern country is replaced by rents and inefficiencies at the level of intermediate-good
producers.5
2.2. Rules of origin in NAFTA
As we explained briefly in the introduction, rules of origin are, in principle, meant to ensure
that goods being exported from one of NAFTA's partners to another truly originate from the
area and are not superficially assembled from components originating from third countries
(henceforth “non-originating”). For that purpose, they specify minimum degrees of
transformation for a good's non-originating inputs in order for that good to qualify for
preferential treatment under NAFTA upon export across the area's internal borders.
The simplest case (rule A)6 concerns goods “wholly obtained or produced entirely in the
territory of one or more of the parties”, which are covered by NAFTA's Article 415. Those
goods, essentially natural resources and agricultural products (although Article 415, reflecting
the wide scope of the negotiators’ concerns, also covers “goods from outer space not
processed in a country other than the parties”) qualify automatically for NAFTA tariff
preference.
Next in order of increasing complication comes rule C covering goods produced with “non-
originating'” materials (i.e. materials imported from non-NAFTA countries) having been first
5
Alternatively, one may take as exogenously given that all intra-bloc tariffs must go to zero, which then implies
that the rate of tariff preference is equal to the rate of MFN tariffs. Then, if MFN tariffs proxy for lobbying
power, ROOs will be positively correlated with them (and with tariff preference since the two are just the same
thing) but for a reason that is unrelated to the Southern member’s participation constraint. We are grateful to a
referee for attracting our attention to this point.
6
On this, see e.g. Giermanski and Sauceda (1994)
9transformed in the NAFTA area. This case covers e.g. furniture made of parts produced in the
area from non-originating steel. Goods produced using such “domestically-transformed”
foreign inputs also enjoy NAFTA tariff preference.
The third case (rule B) covers goods made of foreign, non-transformed inputs. Preferential
treatment for those goods applies only if the rules of origin detailed in NAFTA's Annex 401
are satisfied. Those rules fall under three broad categories, all setting minimum levels of
transformation between the non-originating inputs and the final good being exported across
NAFTA's internal borders. The first type require changes in tariff classification at various
levels of the Harmonized System's nomenclature, with higher levels corresponding to more
substantial transformation. The second type specify a minimum proportion of value added
being realized in the area. The third type specify detailed technical requirements.
More complicated rules (Rule D and a wealth of specific provisions) apply to special cases,
including so-called maquiladoras and the special regime covering textiles and clothing.
“Maquiladoras” is a term referring to production units doing offshore assembly work for the
US market. Generally, they are owned by US companies and at least some of the components
they assemble originate from the US. Maquiladoras enjoyed preferential tariff treatment in the
US before NAFTA was formed. Duties were assessed only on the non-US originating fraction
of the goods' value, which entailed substantial tariff reductions since a large share of the
maquiladoras did assembly work with intermediate products originating from the US. When
NAFTA came into force, the tariff exemption was extended to value added realized in
Mexico.
Maquiladoras have also benefitted from a special regime in the early years of NAFTA, that is,
until 2001. Under Mexico's trade legislation, maquiladoras have enjoyed a form of
preferential treatment frequently encountered in emerging countries under either Export
Processing Zone (EPZ) or Duty-Drawback (DD) schemes. Under the former, imported inputs
enter the country duty-free provided that they are used by firms that legally enjoy EPZ status,
which in general means that the bulk of their output must be for export. Under the latter,
duties paid on imported inputs are refunded to companies who can document that the inputs
have been used for the production of exported goods.
10When a PTA is formed, it is frequently the case that firms exporting into what has become a
preferential area cannot claim DD or EPZ treatment anymore (on this, see Cadot, de Melo and
Olarreaga 2000). By contrast, Mexico's maquiladoras have benefitted from an extension of the
tariff-free status for non-originating inputs well into NAFTA's phase-in period. In fact, the
treatment of maquiladora production under NAFTA was so hassle-free that those who wanted
to sell on Mexico's internal market were said to ship their output across the border to the US
and then back into Mexico in order to avoid paying duties on imported inputs (Vargas 2001).
Starting in 2001 (included), however, NAFTA's Article 303, which allows only limited
refunds of duties paid on inputs,7 applies to non-originating ones.
NAFTA's regime for textiles is particularly complex (see US Customs, 1998). The basic rules
are so-called “yarn forward” and “fiber forward” rules according to which textile and clothing
products are deemed originating provided that they are made of yarn or fiber (whichever
applies) produced in the area. For instance, apparel produced in Mexico is deemed originating
if it is made of non-originating wool fiber (HTS 510105) or cotton fiber (HTS 510103), but
not if it is made of non-originating wool yarn (HTS 510613) or cotton yarn (HTS 520412).
Put differently, apparel products imported into the US must satisfy a “triple transformation”
rule requiring domestic content at each one of three transformation stages: Fiber to yarn, yarn
to fabric, and fabric to garment. Textile and apparel products that do not meet NAFTA's
ROO requirements may nevertheless qualify for NAFTA's preferential tariffs under the
Mexican Special Regime, a program applying for certain goods assembled in Mexico from
US formed and cut components that was especially advantageous in the early years of
NAFTA when the tariff phase-out was incomplete (US Customs 1998). They can also qualify
for reduced tariff rates under Tariff Preference Levels (TPL) up to quantity ceilings (a form of
preferential tariff-quotas) specified in the Agreement's Annex 300b (Schedule 6b). TPL
applies to a few goods outside of textiles and apparel as well.
In this paper, we use a codification of NAFTA’s ROOs performed by Estevadeordal (1999).
The resulting index, which associates a numerical value to every ROO type or combination of
types is as close as one can get to an objective measure of ROO stringency. For changes of
tariff classification (the bulk of NAFTA's ROOs), Estevadeordal's index values are increasing
in the width of the change required. Thus, a change of tariff line has a value of one, a change
11of tariff sub-heading a value of two, and so on. Index values also embody information on
regional value content and technical requirements. The frequency distribution of index values
(by Mexican dollar export values) is shown in Figure 1.
Figure 1
Frequency distribution of Estevadeordal’s ROO restrictiveness index
3. Assessing the costs of NAFTA's ROOs
Tariff preferences granted under NAFTA are substantial. In 1998, about half-way down the
tariff phase-out, the US average tariff on Mexican goods was 0.28%, whereas the US average
MFN tariff on the same goods was 4.8%, giving a preference of 4.51%.8 By 2000, the
respective tariff rates were 0.08% and 4.02%, giving a preference of 3.94%. As a result,
Mexican products can enter the US market at substantially lower tariff rates than competing
products from non-NAFTA countries.9 In 1998, for instance, the tariff on Mexican apparel
products was less than one percent whereas similar products from China and Hong-Kong
faced tariffs of 12.7% and 17.5% respectively. Automobile products had a tariff of 0.4%,
against 2.7% for corresponding products from Germany (on this, see e.g. Lopez-Cordova
2001). In spite of these preferences, however, NAFTA’s utilization rates widely considerably
across sectors and are sometimes substantially below 100% in sectors where tariff preferences
are deeper than average, suggesting that hidden barriers undo at least partially the positive
effect of tariff preference.
7
Specifically, Article 303 allows the refund of the lesser of (i) duties paid in Mexico on imported inputs; or (ii)
duties paid on the final good upon entry in the US or Canada.
8
Average tariffs are ad-valorem equivalents weighted by shares in Mexican exports to the US. Thus, the
weighted-average MFN tariff reported here is computed using the product shares in Mexican exports to the US
as weights, rather than product shares in total US imports.
9
A substantial proportion of the expansion in Mexico's trade with the US in the last decade is due to the
increased activity of maquiladoras, plants located close to Mexico's border with the US many of which do final
assembly work for US-based companies. Maquiladora-type activity is not limited to Mexico. For most
countries, the exports of these companies into the US is registered by US customs under tariff heading 98020050
(“Articles returned to the U.S. after having been exported for repairs or alterations, nesi. Articles returned to the
U.S. after having been exported for repairs or alterations, nesi”). However, for Mexico, maquiladora exports are
registered by US customs under NAFTA. Mexico's exports under tariff heading 98020050 amounted in 2000 to
only $223 million. On this, see Hombeck (1998).
123.1 Utilization rates
As we noted in the introduction, utilization rates give a good indication of the attractiveness
of a preferential regime vis-a-vis MFN treatment. Starting in 2000, customs data on regimes
used by exporters of goods entering the US has been made available by the ITC. Given that
NAFTA’s coverage is close to 100% (i.e., practically all goods are eligible), this data can be
used to construct utilization rates. On that basis, NAFTA's overall utilization rate was, in
2000, 64%, with large fluctuations across sectors and within some of them (see Table 1 and
Figure 2). When tariff lines with no US MFN tariffs (and hence no tariff preferences) are
excluded (last column of Table 1), the overall utilization rate is 83%. As far as large sectors in
Mexico's trade are concerned, the highest utilization rates (with no lines excluded) were 99%
for vegetables (HTS2 chapter 07, $1.6 billion of exports), 97% for vehicles (HTS2 87, $26
billion), 85% for plastics (HTS2 39-40, $1.8 billion), 80% for footwear (HTS2 64-67, $414
million), and 72% for steel (HTS2 72 and 73, $700 million). At 66%, the textile and apparel
sector (HTS2 chapters 50-63, $10.3 billion) had a barely higher-than average utilization rate.
At the other end of the spectrum, the lowest utilization rates among significant sectors were
20% for furniture (HTS chapter 94, $3.8 billion), 42% for optics (HTS2 90,$4.6 billion), 48%
for machinery (HTS2 84-85,$ 53 billion), and 48% for knitting products (HTS2 61,$ 3.5
billion), the lowest rate in the textile-clothing sector. The results for textile and clothing
products are particularly striking given that in 2000, the ad-valorem equivalent of the US's
average MFN duty (weighted at the 6-digit level by the value of Mexican exports) on HTS
section 11 was 16.7%, whereas the ad-valorem equivalent of the rates applicable under
NAFTA was practically zero.
Table 1
Mexican exports to the US and NAFTA’s regime
Note: the number of tariff lines with utilization rates strictly between zero and one is 1471 and the corresponding
value of imports is $98.3 billion or 75.6% of the total.
Figure 2
NAFTA utilization rates, 2000
Relatively low and unequal utilization rates in the face of sometimes substantial tariff
preference call for an explanation. For example, transport equipment sector (HTS17) and
leather goods (HTS8) have similar tariff preferences (6.28 and 6.38 percent respectively) but
13their NAFTA utilization rates are very different (94.9 versus 57 percent). The most likely
explanation lies in differences in the cost of complying with ROOs, as Estevadeordal’s index
value is 4.8 (out of 7) for transport equipment but 5.6 for leather goods. We now turn to an
assessment of the quantitative importance of ROOs and other administrative costs.
3.2 A “revealed-preference” index
As briefly explained in the introduction, in this section we approach the determination of the
cost of ROOs through an indirect “revealed-preference” method. For those tariff headings that
are eligible for NAFTA treatment and where 100% of Mexico's exports to the US enter under
NAFTA regime, the combined cost of complying with ROOs and other NAFTA-related
administrative procedures is no greater than the benefit conferred by preferential tariff access.
In other words, the value of b in (9) is positive. For those tariff headings, thus, the rate of
preference gives an upper bound on combined ROO-administrative costs. By the same
reasoning, for those tariff headings where none of Mexico's exports enter the US under
NAFTA regime, the rate of preference gives a lower bound on those costs. Finally for all
those tariff headings in between, as explained above one can suppose that, at least for the
marginal firms (administrative and ROO compliance costs may vary across firms), b = 0.
Thus, for those tariff headings, the rate of preference gives an approximation of combined
costs. Eliminating tariff headings with either 100% or none of the shipments under NAFTA,
the import-weighted average of the rate of preference is 5.06% (1471 observations, standard
error 0.58%). This gives a first approximation on combined ROO and administrative costs
associated with the preferential regime. We now propose a method for disentangling those
costs.
As in the case of ROOs, getting direct estimates of non-ROO administrative costs is difficult,
and we use a roundabout method. Those costs can be expected to vary across firms,
depending on their administrative capabilities, but not necessarily to vary systematically
across sectors. Proceeding on the assumption that they are roughly the same across sectors,
we combine Estevadeordal's ROO index and the revealed-preference approach described
above. Let U be the set of tariff headings (at the 6-digit level) in which more than a given
percentage u* of Mexican exports enter under NAFTA regime. That is, U = {i : ui > u*}. As
explained in the previous section, for tariff headings, if u* is sufficiently close to 100% the
14rate of tariff preference gives an upper bound on combined ROO and administrtative costs. In
order to make sure that we catch only industries where ROO costs are likely to be low, let
also R be the set of tariff headings for which Estevadeordal's ROO index is no more than a
cutoff r*. Table 1bis reports the average, minimum and maximum values of tariff preferences
in three different sets R∩U corresponding to three different cutoffs u*: 90%, 95% and 99%.
The only value used for r* is two (change of tariff heading) because there is no tariff line with
ri = 1 and ui > 90%.
Table 1bis
Tariff preferences in R∩U
Minimum values are nil, but average values are between 3% and 3.5% and robust to changes
in u*. As a further check, Figure 3 shows the frequency distribution of tariff preference levels
by intervals of two percentage points in R∩U (218 observations) for u* = 95% and confirms
that the range 2-4% is not only the average but also the mode of the distribution. The average
tariff-preference level for u* = 99% (3.12%) can thus be taken as the closest upper bound on
non-ROO administrative costs. It is likely to be over the true value given that it is quite high
compared to other estimates of the bookkeeping costs associated with PTAs and implies very
low ROO costs (5.06 - 3.12 = 1.94%).
Figure 3
Frequency distribution of tariff preference levels in R∩U
3.3 Tariff preference and ROOs: offsetting instruments?
This section explores the combined effects of NAFTA’s tariff preference and ROOs on the
direction of Mexican exports. Data on US imports from Mexico (overall and under NAFTA
regime) at the HTS 6-digit levels are taken from the ITC. Ad-valorem tariff equivalents were
computed at the 8-digit level using US ad-valorem and specific tariffs under MFN and
NAFTA regimes and aggregated to the 6-digit level using US imports from Mexico as
weights.10 As a starting point, Table 2 below shows that there has been limited change in the
sectoral composition of Mexico’s exports to the US between 1992 and 2000 —essentially in
sectors like computers, electrical machinery and transport equipment where maquiladoras
15have developed. In sensitive sectors like textiles, apparel, footwear, food and steel, the
changes have been marginal at best.
Table 2
Sectoral composition of Mexico’s trade with the US, 1992-2000
Figure 4 shows the location of Mexican exports on a tariff preference/ROO space at the sector
level (HTS1). The size of dots is proportional to each sector's share in Mexico's total exports
to the US. Sectors lying to the Northwest of the picture have large tariff preferences and low
ROOs and are therefore potential winners in terms of market access. In other words, if tariff
preference is taken as a “good” and ROOs as a “bad” (both defined strictly in terms of
market-access improvement, leaving welfare considerations aside), the attractiveness of the
high-preference, high-ROO package offered by NAFTA to HTS 11 (textiles) cannot be
directly compared with the low-preference, low-ROO package offered to HTS 16 (machinery
and electrical equipment), but the latter is unambiguously better than that offered to HTS 2
(vegetable products) and probably better than that offered to HTS15 (base metals). This
picture provides a benchmark to explore the implications of NAFTA-like treatment for other
LA countries. Figure 4 clearly shows a 'frontier' in terms of tariff preferences and ROOs with
all points lying Southeast of that frontier. The picture is thus consistent with the notion of a
binding participation constraint for Mexico.
Figure 4
The location of Mexican exports in PREF/ROO space
Given that there is substantial variation in NAFTA's tariff preference across tariff lines, in the
absence of offsetting administrative or ROO costs one would expect Mexico's trade flows to
be affected by NAFTA's rate of preference. Specifically, one would expect to observe, for
items with deep preference under NAFTA, a higher share in Mexico's exports to the US than
in Mexico's exports to other markets. By contrast, if the cost of complying with NAFTA's
ROO and other administrative hassle offsets the benefit of tariff preference, one would expect
the composition of Mexico's trade flows to be largely unaffected by NAFTA tariff preference.
That is, under the latter hypothesis and provided that the pattern of US MFN tariffs does not
differ too strongly from the pattern of MFN tariffs applied by Mexico's other trading partners,
10
Specific tariffs were converted into ad-valorem equivalents using the unit values of imports as prices.
16one should not be able to trace large differences between the pattern of Mexican exports to the
US vs. to the rest of the world. Thus, comparing Mexico's exports to the US and to the world
provides a further check on the hypothesis that NAFTA involved a switch of protection
instrument from tariff to ROOs rather than an overall reduction in the level of protection.
We explored this question by estimating the following equation using WLS with Mexican
exports as weights at the six-digit level for the year 2000:
XUSi = α0 + α1 * XROWi + α2 * ln PREFi + α3 * ln ROOi + Σkαk * Dki (10)
where XUSi stands for Mexico’s exports to the US in tariff line i, XROWi is Mexico’s exports
elsewhere than to the US, PREFi stands for the rate of tariff preference under NAFTA, ROOi
is Estevadeordal’s index, and Dki is a vector of dummy variables by HTS chapters. The log
forms for PREF and ROO give a better fit than linear form. They imply that, ceteris paribus,
XUSi is an increasing, convex function of PREFi if α1 is positive, and a decreasing, convex
function of ROOi if α3 is negative as ∂2 XUSi /∂ (ROOi)2 = − α3 / (ROOi)2 > 0 . Tariff lines
with PREF = 0 were assigned a value arbitrarily close to zero (10E-13).11
We also estimated a variant of (10) in which Estevadeordal’s index was replaced by a vector
of dummies for specific forms of ROOs. Let R = (HEAD, SUBHEAD, ITEM, EXC, RVC)
where
HEAD = 1 when the ROO requires a change in tariff heading
SUBHEAD = 1 when it requires a change in tariff sub-heading
ITEM = 1 when it requires a change in tariff item
EXC = 1 when there is one or more exceptions
RVC = 1 when the ROO specifies a minimum regional value content.
Some tariff lines have ROOs expressed as required changes of tariff chapter, but the CHAP
variable comes out with the wrong sign (i.e. positive) except when interacted with dummy
variables for food and for textiles (where most ROOs take the form of a change of chapter), in
which case it has the expected negative sign although not significantly different from zero.
This is not overly surprising given that changes of tariff chapter correspond (by construction)
to high values of Estevadeordal’s index for which the functional form in (10) implies weaker
11
Taking them out of the sample reduces the number of observations by more than a thousand and somehow
reduces the precision of estimates.
17effects, which may not be estimated with sufficient precision. Accordingly we report in Table
3 below the results of a specification with interaction terms between CHAP and FOOD and
TEXTILE. The equation to be estimated is then
XUSi = α0 + α1 * XROWi + α2 * PREFi + α3 * Ri + α4 * CHAP * FOOD
+ α5 * CHAP * TEXTILE + Σkαk * Dki (11)
where α2 is the vector of coefficients on the components of Ri.
Before we discuss the results, two issues must be dealt with. First, if PREF and ROO are
substitutes, there may be collinearity (in a weak sense) between the two. However regressing
ROO on PREF gives a positive and significant parameter estimate (consistent with
substitutability) but an R2 of only 10%, suggesting that the association is not sufficiently close
as to be a problem in the estimation of (10). Second, it can be argued that ROO and PREF are
endogenous to Mexican exports if tariff and ROO protection are used to restrict Mexican
access to the US market. However ROOs determined in the course of negotiations held in the
early 1990s and finalized in 1992 can hardly be endogenous to Mexico’s 2000 export
pattern.12 As for PREF, GATT Article XXIV implies that intra-bloc tariffs have to go to zero,
so steady-state tariff preferences are equal to MFN tariffs which are also predetermined (see
footnote 4 supra). Estimation results are shown in Table 3.
Table 3
Regression results for (10) and (11)
The results from (10) are as expected. The relationship between exports to the US and exports
to the rest of the world is proportional with a factor between three and four, but tariff
preference has a positive influence on Mexico's exports to the US. ROOs have the opposite
effect, and both are highly significant. We will comment below on the magnitude of marginal
effects.
12
Technically, the ROO variable can be considered as predetermined to the dependent variable, which implies
that there is no correlation between the regressors and the equation’s error term, hence that OLS and WLS
estimates are unbiased. An equation determining ROOs on the basis of contemporaneous variables can be found
in Estevadeordal (1999), but simultaneous estimation of these two in a recursive system would not alter the point
estimates of (10).
18The results of (11), in which Estevadeordal’s index is decomposed into dummy variables for
various types of ROOs, are also quite interesting. Changes of tariff classification have
negative and significant effects, whereas exceptions have positive effects. This suggest that
the bulk of exceptions to ROOs make them less constraining rather than more, unlike the oft-
cited restriction on tomato paste according to which ketchup is deemed originating if it results
from a transformation of ingredients satisfying a change-of-chapter rule, but not if it results
from the transformation of tomato paste (see Krueger 1999). Regional value content rules
appear particularly significant and have large marginal effects.
In order to estimate the quantitative effects of each instrument on the direction of Mexican
trade flows, we performed the following exercise. On the basis of the parameter estimates in
(10) and (11), we compared the predicted values of Mexican exports to the US in three cases:
(i) with actual values of the PREF and ROO variables (NAFTA as it is, i.e. the benchmark
case); (ii) with no tariff preferences and no rules of origin,13 which we interpret as “no
NAFTA”; and (iii) with NAFTA tariff preferences but no ROOs (a hypothetical NAFTA
without rules of origin). The difference between case (i) and case (ii) gives an estimate of the
direct effect of NAFTA's package (tariff preference and ROOs) on Mexican trade flows.
Results are presented as percentage deviations from the relevant equation’s baseline predicted
value for Mexico’s exports to the US, namely is $152.3 billion with (10) and $133.4 billion
with (11). The results are shown in Table 4.
Table 4
Simulation results for (10) and (11)
Consider the first part of Table 4, based on (10). If “No NAFTA” is interpreted as setting
ROOs at their lowest level, then the combined effect of tariff preferences and ROOs
(NAFTA’s package) raises Mexican exports, on average, by only 3.1%. As “No NAFTA” is
interpreted as elimination of tariff preference but ROOs set at higher levels, NAFTA’s effect
appears more favourable. As explained above, setting ROOs to their lowest level is the most
13
The exercise we perform is as follows. In case (i), we use actual values of the PREF and ROO variables to
predict the value of Mexico’s exports to the US. In case (ii), we set PREF equal to 10E-13 across the board and
ROO to a ‘low’ value across the board. The first part of Table 4 reports results for three values of ROO: 1, 2 or
3. The reason for not setting the ROO variable to zero is that, under NAFTA, there is no tariff line with ROO
equal to zero, so that predicting the value of XUS (the dependent variable) so far out of the sample with non-
linear forms gives unreasonable results. Results based on setting ROO equal to higher values are more
conservative but arguably less prone to prediction errors. If anything, the bias that this introduces reinforces the
point we are making, since setting ROOs at a lower level would generate larger negative effects.
19logical experiment to perform but stretches the predictive power of the equation to the
sample’s bounds, which may induce errors. With this caveat in mind, it is fair to say that the
marginal effects of tariff preferences and ROOs as they are in NAFTA’s present form seem to
produce limited positive net effects (+11.7% with ROO=2 taken as the “No NAFTA” value).
The second column shows that if tariff preferences were maintained but ROOs eliminated the
positive effects on Mexico’s exports would be considerable (+35.3% if ROOs were set across
the board at a level corresponding to ROO=2).
The second part of the table provides further indications on the effects of different types of
ROOs. As for changes of tariff classification, the most common type of ROOs, note that
relaxing ITEM (changes of tariff item), which has the largest marginal effects in (11),
produces only a minor effect on trade flows as this type of ROO affects only low-volume
tariff lines. Conversely, relaxing CHAP which has a low and imprecisely-estimated marginal
effect produces a large change on textile and food exports. Relaxing HEAD (change of tariff
heading) also produces a dramatic effect on Mexican trade flows.
Several caveats are in point. First, the exercise cannot measure non-trade effects of NAFTA
(e.g. on the credibility of reforms) and should therefore be taken as a lower bound on
NAFTA's real-world effect. Second, our results are based on effects measured on a cross-
sectional data set and cannot give a full picture of NAFTA's effects since effects that cut
across all sectors effects are subsumed in the constant. Thus, at least one important question
remains unsanswered: namely, whether the recent expansion of Mexico's exports to the US is
indeed attributable to NAFTA but to effects that are only indirectly related to tariff
preferences, or whether it is attributable instead to exchange-rate or other macroeconomic
effects. Even more importantly, the effect of preferential treatment may be to generate inward
FDI flows in sectors that benefit from substantially improved market access. This type of
effect cannot be assessed on the basis of a “snapshot” of Mexico's export composition, but
require a panel estimation over a sufficient number of years.
With these caveats in mind, the provisional conclusion here is that, at least at first sight,
Mexico's export pattern seems to have been affected positively but in a quantitatively small
way by the combined effect of NAFTA's tariff preferences and ROOs, because the negative
effect of the latter largely offsets the positive effect of the former. This has two implications.
First, it provides renewed support for the view that the gains from tariff liberalization under
20preferential trading agreements can be largely offset by non-tariff compliance costs (in which
case PTAs involve a substitution of instruments rather than the simple elimination of one of
them). Second, the substitutability between tariff and ROO protection means that NAFTA has
so far not created a very clear pattern of winners and losers among Mexican exporters. This
provides a useful starting point for analysing the potential effect of extending NAFTA
treatment to other Latin American countries under the FTAA initiative, since evenly spread
benefits across industries imply (as a first approximation) evenly spread benefits across
countries as well.
4. FTAA Implications
We now turn to an extension of the analysis to other Latin American (LA) countries, the
question being the extent to which LA countries would benefit from trade preferences granted
through the FTAA proposal on the agenda of the current US administration. Our analysis is
perforce largely based on conjectures, and several caveats are in point. First, our approach
consists of comparing current LA trade patterns with NAFTA’s pattern of trade preferences
and ROOs. This entails several potential sources of bias. First, the opening of talks with
Mercosur may create a window of opportunity for renegotiating NAFTA’s terms in a way that
would make them more favourable to Southern members. Second, as companies located in
eligible countries anticipate having one day to comply with NAFTA’s ROOs, they may
redirect their sourcing policies, making LA trade patterns more “NAFTA-compatible” than
they currently appear. However, none of this source of bias is likely to alter substantially our
conclusions. As to the first, if some adjustment in NAFTA’s rules is foreseeable, it seems
unlikely that a wholesale renegotiation would take place to accommodate new partners.14 As
to the second, prior adjustment of trade patterns in anticipation of the need to comply with
ROOs is endogenous to the trade agreement and it is indeed correct to look at trade patterns
before such adjustment has taken place. More seriously, as in the case of Mexico, a full
analysis of the impact of ROOs on market access would require input-output data which is not
available at sufficiently disaggregated levels and in trade (HS or SITC) classification.
Therefore we perform the analysis in two steps, looking first at export patterns and then at
import patterns, distinguishing between intermediate and final products.
14
Indeed, NAFTA’s formation did not involve substantial renegotiation of the Canada-US FTA. In particular,
sensitive sectors were left untouched (see Wonnacott, 1996).
214.1 Regional preferences and distance
As a preliminary remark, note that the Western Hemisphere’s geography means that there is a
big difference in terms of distance to the main market between Mexico and eligible countries
in South America. Table 5 shows the ad-valorem equivalent of transportation costs between
the US and a number of LA countries. Distances vary considerably, and as suggested by the
literature on “natural trading partners”, the effect of trade preference is likely to be eroded by
distance. Indeed, the relocation of a large chunk of Mexico’s export-oriented manufacturing
activity close to the US border shows that proximity matters.15
Table 5
AVE of transportation costs to the US, LA countries, 1994
Table 5 shows that Chile, with transportation costs to and from the US equivalent to an ad-
valorem tariff of 12%, is hardly in a position to benefit dramatically from tariff reduction
from a baseline MFN rate of 4%. As a matter of fact, only in Mexico's case is the AVE of
transportation costs to the US lower than the average rage of NAFTA's tariff preference.
How distance interacts with the effects of ROOs is a more subtle question it looks. On one
hand, the cost of the constraints imposed by a given set of ROOs on the sourcing policies of
intra-bloc exporters increases with distance from the intra-bloc source of intermediate inputs,
especially if those are costly to transport.16 On the other hand, in some special cases trade
diversion replacing tariff revenue by intra-bloc transportation costs (when the latter are just
below the preference margin) can be eliminated by ROOs, which are then welfare-enhancing
even though they reduce intra-bloc market access (see Krishna and Krueger 1995, footnote 4).
However, the power of this argument is reduced by the high level of transportation costs
noted above. Thus, the scope for substantial market-access improvement appears very limited
15
On NAFTA’s differential effects on Mexico’s regions, see Esquivel, Lederman, Messmacher and Villoro
(2002).
16
We are grateful to an anonymous referee for attracting our attention to this point. However, it should be noted
that an integrated PTA like the FTAA is better in this regard than a hub-and-spoke system (except if it has
cumulation of ROOs like the Europe Agreements with the CEECs). In other words, the argument is weakened if
Argentine companies can satisfy FTAA ROOs by sourcing in Brazil instead of just in the US.
22at the aggregate level. Only for sectors with substantially lower-than average transportation
costs can NAFTA-like tariff preference make a difference.
4.2 LA export patterns
Notwithstanding the distance issue, whether the results of the last section —namely, that
Mexico's market access to the US was only marginally improved by NAFTA tariff reductions
if one takes into account the implicit cost of ROOs— carry over to other LA countries
potentially eligible for preferential treatment under the FTAA depends, inter alia, on a
combination of two things. If FTAA treatment involves primarily the replication of NAFTA's
pattern of tariff preference and ROOs, and LA trade patterns (in terms of geographical and
input/output composition) are correlated with Mexico's, results for Mexico can be expected to
carry over to those countries. In other cases, FTAA effects are more uncertain. As mentioned
earlier, how close FTAA treatment would be to NAFTA remains to be seen, since
negotiations have not yet started except for Chile (for which they were nearly complete at the
time of writing).
What can be assessed, first, is how close are the export patterns of Mexico and other LA
countries. Tables 6 and 7 show that LA exports to the US are all several orders of magnitude
smaller than Mexico's (see Soloaga and Winters 1999 for gravity estimates). So even in the
hypothetical case of substantial market-access gains the volumes involved are likely to be
small.
Table 6
Sectoral composition of LA exports to the US ($million)
Table 7
US and NAFTA shares in LA exports
Table 8(a) and 8(b)
Cross-correlations of LA exports
The cross-correlation of LA exports is given in Tables 8(a) and 8(b). The first observation is
that, once oil is taken out, the correlations are all below 0.5 (only the correlation between
23Ecuador and Columbia being close to 0.5). Thus, any conclusion applying to Mexico should
be taken to other LA countries only carefully.
Next, we have drawn in Figures 5(a)-5(g) the equivalent of Figure 4 for LA countries other
than Mexico. The size of dots is proportional to each sector's share in the country's total
exports and is thus country-specific, whereas their location is determined by the NAFTA
Treaty and is thus common to all countries. Countries with larger dots close to the implicit
'frontier' stand to gain more from NAFTA than others. Figures 5(a)-5(g) suggest that Brazil
and Costa Rica stand to gain more substantially than other LA countries from NAFTA
treatment. Brazil has large exports in HTS16 (machinery and electrical equipment), HTS17
(transportation equipment) and 11 (textile and apparel) all of which lie along the Northwest
frontier of the picture. Costa Rica also has substantial exports (relative to its own total) in
HTS11 and 16.
For other LA countries, the prospects are less encouraging. Venezuela exports essentially oil,
which has very low tariff preference since US MFN tariffs on oil are low to start with. Chile,
which exports essentially copper, live animals (HTS1),17 vegetable products (HTS2), and
wood products (HTS9) stands to gain only in a limited way from NAFTA treatment unless the
cross-sectoral pattern of ROOs is substantially renegotiated.18 Limited gains also seem to be
predictable for Ecuador given the low level of their manufactured products exports. Argentina
has a more diversified manufacturing and export base, but its exports under HTS 16 and 17
(electrical machinery and transport equipments respectively) are limited.
In sum, given that Mexico, which under NAFTA's mixture of tariff preference and ROOs has
a favorable export pattern (at least ex post) relative to other LA countries, seems to have
enjoyed fairly limited market-access improvements, the prospects for dramatic improvements
are unlikely for other LA countries. Of course, these preliminary results are based on static
measures of the specialization of exports. Improved market access may also induce changes in
the export specialization of beneficiary countries. If such is the case, NAFTA’s benefits do
17
It is not entirely clear however in the absence of in-depth case studies to assess whether ROOs on live animals
are really binding on Chile given the high transportation costs that the country faces which in any case limit its
participation in internationally integrated cattle ranching systems.
18
At the time of writing, negotiations on a preferential trading agreement between Chile and the US were largely
complete but not yet in the public domain.
24You can also read
